Leech

Let's say you want to read some sort of fiction. You're a fan of it, perhaps. But mobile websites are kind of non-ideal, so you'd like a proper ebook made from whatever you're reading.

Setup

You need Python 3.9+ and poetry.

My recommended setup process is:

$ pip install poetry
$ poetry install
$ poetry shell

...adjust as needed. Just make sure the dependencies from pyproject.toml get installed somehow.

Usage

Basic

$ python3 leech.py [[URL]]

A new file will appear named Title of the Story.epub.

This is equivalent to the slightly longer

$ python3 leech.py download [[URL]]

Flushing the cache

$ python3 leech.py flush

Learn about other options

$ python3 leech.py --help

If you want to put an ePub on a Kindle you'll have to either use Amazon's send-to-kindle tools or convert it. For the latter I'd recommend Calibre, though you could also try using kindlegen directly.

Supports

• Fanfiction.net
• FictionPress
• ArchiveOfOurOwn
  • Yes, it has its own built-in EPUB export, but the formatting is horrible
• Various XenForo-based sites: SpaceBattles and SufficientVelocity, most notably
• RoyalRoad
• Fiction.live (Anonkun)
• DeviantArt galleries/collections
• Sta.sh
• Completely arbitrary sites, with a bit more work (see below)

Configuration

A very small amount of configuration is possible by creating a file called leech.json in the project directory. Currently you can define login information for sites that support it, and some options for book covers.

Example:

{
    "logins": {
        "QuestionableQuesting": ["username", "password"]
    },
    "images": {
        "image_fetch": true,
        "image_format": "png",
        "compress_images": true,
        "max_image_size": 100000,
        "always_convert_images": true
    },
    "cover": {
        "fontname": "Comic Sans MS",
        "fontsize": 30,
        "bgcolor": [20, 120, 20],
        "textcolor": [180, 20, 180],
        "cover_url": "https://website.com/image.png"
    },
    "output_dir": "/tmp/ebooks",
    "site_options": {
        "RoyalRoad": {
            "output_dir": "/tmp/litrpg_isekai_trash",
            "image_fetch": false
        }
    }
}

Note: The image_fetch key is a boolean and can only be true or false. Booleans in JSON are written in lowercase. If it is false, Leech will not download any images. Leech will also ignore the image_format key if images is false.

Note: If the image_format key does not exist, Leech will default to jpeg. The three image formats are jpeg, png, and gif. The image_format key is case-insensitive.

Note: The compress_images key tells Leech to compress images. This is only supported for jpeg and png images. This also goes hand-in-hand with the max_image_size key. If the compress_images key is true but there's no max_image_size key, Leech will compress the image to a size less than 1MB (1000000 bytes). If the max_image_size key is present, Leech will compress the image to a size less than the value of the max_image_size key. The max_image_size key is in bytes. If compress_images is false, Leech will ignore the max_image_size key.

Warning: Compressing images might make Leech take a lot longer to download images.

Warning: Compressing images might make the image quality worse.

Warning: max_image_size is not a hard limit. Leech will try to compress the image to the size of the max_image_size key, but Leech might not be able to compress the image to the exact size of the max_image_size key.

Warning: max_image_size should not be too small. For instance, if you set max_image_size to 1000, Leech will probably not be able to compress the image to 1000 bytes. If you set max_image_size to 1000000, Leech will probably be able to compress the image to 1000000 bytes.

Warning: Leech will not compress GIFs, that might damage the animation.

Note: if always_convert_images is true, Leech will convert all non-GIF images to the specified image_format.

Arbitrary Sites

If you want to just download a one-off story from a site, you can create a definition file to describe it. This requires investigation and understanding of things like CSS selectors, which may take some trial and error.

Example practical.json:

{
    "url": "https://practicalguidetoevil.wordpress.com/table-of-contents/",
    "title": "A Practical Guide To Evil: Book 1",
    "author": "erraticerrata",
    "chapter_selector": "#main .entry-content > ul:nth-of-type(1) > li > a",
    "content_selector": "#main .entry-content",
    "filter_selector": ".sharedaddy, .wpcnt, style",
    "cover_url": "https://gitlab.com/Mikescher2/A-Practical-Guide-To-Evil-Lyx/raw/master/APGTE_1/APGTE_front.png"
}

Run as:

$ ./leech.py practical.json

This tells leech to load url, follow the links described by chapter_selector, extract the content from those pages as described by content_selector, and remove any content from that which matches filter_selector. Optionally, cover_url will replace the default cover with the image of your choice.

If chapter_selector isn't given, it'll create a single-chapter book by applying content_selector to url.

This is a fairly viable way to extract a story from, say, a random Wordpress installation with a convenient table of contents. It's relatively likely to get you at least most of the way to the ebook you want, with maybe some manual editing needed.

A more advanced example with JSON would be:

{
    "url": "https://practicalguidetoevil.wordpress.com/2015/03/25/prologue/",
    "title": "A Practical Guide To Evil: Book 1",
    "author": "erraticerrata",
    "content_selector": "#main .entry-wrapper",
    "content_title_selector": "h1.entry-title",
    "content_text_selector": ".entry-content",
    "filter_selector": ".sharedaddy, .wpcnt, style",
    "next_selector": "a[rel=\"next\"]:not([href*=\"prologue\"])",
    "cover_url": "https://gitlab.com/Mikescher2/A-Practical-Guide-To-Evil-Lyx/raw/master/APGTE_1/APGTE_front.png"
}

Because there's no chapter_selector here, leech will keep on looking for a link which it can find with next_selector and following that link. We also see more advanced metadata acquisition here, with content_title_selector and content_text_selector being used to find specific elements from within the content.

If multiple matches for content_selector are found, leech will assume multiple chapters are present on one page, and will handle that. If you find a story that you want on a site which has all the chapters in the right order and next-page links, this is a notably efficient way to download it. See examples/dungeonkeeperami.json for this being used.

If you need more advanced behavior, consider looking at...

Adding new site handlers

To add support for a new site, create a file in the sites directory that implements the Site interface. Take a look at ao3.py for a minimal example of what you have to do.

Images support

Leech creates EPUB 2.01 files, which means that Leech can only save images in the following format:

• JPEG (JPG/JFIF)
• PNG
• GIF

See the Open Publication Structure (OPS) 2.0.1 for more information.

Leech can not save images in SVG because it is not supported by Pillow.

Leech uses Pillow for image manipulation and conversion. If you want to use a different image format, you can install the required dependencies for Pillow and you will probably have to tinker with Leech. See the Pillow documentation for more information.

To configure image support, you will need to create a file called leech.json. See the section below for more information.

Docker

You can build the project's Docker container like this:

docker build . -t kemayo/leech:snapshot

The container's entrypoint runs leech directly and sets the current working directory to /work, so you can mount any directory there:

docker run -it --rm -v ${DIR}:/work kemayo/leech:snapshot download [[URL]]

Contributing

If you submit a pull request to add support for another reasonably-general-purpose site, I will nigh-certainly accept it.

Run EpubCheck on epubs you generate to make sure they're not breaking.